Int J Adv Manuf Technol (2012) 58:9–17

DOI 10.1007/s00170-011-3365-8

ORIGINAL ARTICLE

Multi-objective optimization of green sand mould system

using evolutionary algorithms
B. Surekha & Lalith K. Kaushik & Abhishek K. Panduy &
Pandu R. Vundavilli & Mahesh B. Parappagoudar

Received: 2 May 2010 / Accepted: 25 April 2011 / Published online: 7 May 2011
# Springer-Verlag London Limited 2011

Abstract The quality of cast products in green sand made to form a single objective, after considering all the
moulds is largely influenced by the mould properties, such four individual objectives, to obtain a compromise solution,
as green compression strength, permeability, hardness and which satisfies all the four objectives. The results of this
others, which depend on the input (process) parameters study show a good agreement with the experimental results.
(that is, grain fineness number, percentage of clay,
percentage of water and number of strokes). This paper Keywords Green sand mould system . Optimization .
presents multi-objective optimization of green sand mould Genetic algorithm . Particle swarm optimization
system using evolutionary algorithms, such as genetic
algorithm (GA) and particle swarm optimization (PSO). In
this study, non-linear regression equations developed 1 Introduction
between the control factors (process parameters) and
responses like green compression strength, permeability, During moulding process, the quality of the parts produced
hardness and bulk density have been considered for depends on the properties (that is, green compression
optimization utilizing GA and PSO. As the green sand strength, permeability, hardness and bulk density) of
mould system contains four objectives, an attempt is being moulding sand. It is important to note that improper levels
of these properties leads to common casting defects, such as
B. Surekha : P. R. Vundavilli (*) blow holes, pinhole porosity, poor surface finish, dimen-
Department of Mechanical Engineering, sional variation, scabs and rat tails, misruns, etc. It is also
DVR & Dr. HS MIC College of Technology,
important to note that the mould properties are influenced
Kanchikacherla, Andhra Pradesh 521180, India
e-mail: [email protected] by a large number of controllable parameters (that is, grain
fineness number, percentage of clay, percentage of water
B. Surekha
e-mail: [email protected] and number of strokes). Hence, it is important to identify
the levels of the input variables that provide required mould
L. K. Kaushik : A. K. Panduy properties, which improves the quality of the parts
Department of Mechanical Engineering,
produced by this mould.
Rungta College of Engineering & Technology,
Bhilai, Chattisgarh 490024, India Most of the research work on moulding sand during
1960s and 1970s was based on experimental and theoretical
L. K. Kaushik
e-mail: [email protected] approaches. The relationship between permeability and
transformation zones, mould pressure, void space control,
A. K. Panduy
e-mail: [email protected] etc., was developed by Marek [1] through substantial
mathematical equations. In addition to this, Frost and Hiller
M. B. Parappagoudar [2] established the pressure and hardness distributions in
Department of Mechanical Engineering,
sand moulds. Later on, Wenninger [3] utilized the rigid
Chhatrapati Shivaji Institute of Technology,
Durg, Chattisgarh 491001, India water theory to explain sand–clay–water relationships. This
e-mail: [email protected] approach was completely theoretical and not supported by a
10 Int J Adv Manuf Technol (2012) 58:9–17

large number of experiments. Moreover, statistical design second approach is based on generating Pareto optimal
of experiments (DOE) had been used by various inves- solution sets that are non-dominated with respect to each
tigators to study the effects of different variables on the other. The present research falls in to the first category.
green sand mould properties. In [4], DOE technique was Evolutionary optimization approaches, such as GA and
applied to study the effect of process variables on bulk PSO, have attracted a great deal of attention in recent
density and green compression strength. In addition to these times. With their better global search abilities, these
approaches, Casalino et al. [5] utilized Taguchi technique to optimization approaches can find global optima more
establish third order model for permeability and compres- quickly through cooperation and competition among the
sion strength in laser sintered sand moulds. Moreover, population of potential solutions. It is important to note
Parappagoudar et al. [6, 7] developed linear and non-linear that GA was used to solve multi-objective optimization
statistical models utilizing full factorial DOE, central problems related to grinding [10], turning [11, 12],
composite design (CCD) and Box-Bhenken design. In the abrasive flow machining [13], wire electric discharge
above work, the authors had considered grain fineness machining [14], drilling and riveting sequence planning
number, percentage of clay, percentage of water and [15], etc. Similarly, PSO was also used to solve the
number of strokes as input parameters and green compres- problems related to grinding [10], electrochemical machining
sion strength, permeability, hardness and bulk density as the [16], steam temperature control [17] and some other
responses. Among the non-linear regression equations engineering problems [18]. To the best of the authors’
developed by the abovementioned three approaches, knowledge, not much work had been reported in the field of
CCD-based model was found to be the more accurate multi-objective optimization of process parameters of green
model for prediction of the responses. Later on, the sand mould systems.
optimization of process parameters of green sand casting In the present paper, the non-linear regression equations
was established in [8] utilizing Taguchi parameter design. developed in [7] has been considered for multi-objective
The process parameters such as green compression optimization, utilizing the most popular evolutionary
strength, moisture content, pouring temperature and mould algorithms, such as GA and PSO. Green compression
hardness vertical and horizontal were considered to identify strength, permeability, hardness and bulk density are
the effect of these parameters on casting defects. As the considered as responses (that is, objectives) and grain
above developed method involved a traditional optimiza- fineness number, percentage of clay, percentage of water
tion method, the solutions obtained were not global optimal and number of strokes are treated as inputs (that is, process
in nature. Therefore, a global optimization method is variables). A single objective has been formed after
required to identify the optimal combination of parameters combining the four responses. Both the GA and PSO
for achieving the desired performance of the green sand algorithms are used to optimize this single objective to
mould system. In single objective optimization, one obtain a solution. It is interesting to note that the results
attempts to obtain the best design or decision, which is obtained by these algorithms are comparable.
usually the global maximum or minimum depending on the The rest of the manuscript is organized as follows.
optimization problem. In green sand mould system, it is Section 2 deals with the formulation of the problem. Tools
difficult to find a single optimal combination of parameters and techniques used in this study are explained in Section 3.
for green compression strength, permeability, hardness and The results are presented and discussed in Section 4. Some
bulk density. Hence, there is a need for a multi-objective concluding remarks are made in Section 5.
optimization method to arrive at the solutions to this
problem. This multi-objective optimization problem can
be converted to a single objective problem after applying a 2 Formulation of the problem
suitable method.
This type of problems can be best solved by utilizing The quality of the parts produced in green sand mould
evolutionary algorithms, such as genetic algorithms (GA), system mainly depends on the properties (responses) of
particle swarm optimization (PSO), etc. The early use of the mould, such as green compression strength (GCS),
evolutionary search was first reported in the 1960s by permeability (P), hardness (H) and bulk density (BD),
Rosenberg [9]. Since then, there had been a growing which in turn depends on the process variables (that is,
interest in devising different evolutionary algorithms for grain fineness number, percentage of clay, percentage of
multi-objective optimization. There exist two general water and number of strokes). Figure 1 shows the
approaches to solve the multi-objective optimization prob- schematic diagram of green sand mould system as an
lem. The first approach deals with combining individual input–output model.
objective functions into a single composite objective The ranges of the process variables used in this study are
function after assigning weight to each objective. The given in Table 1.
Int J Adv Manuf Technol (2012) 58:9–17 11

Fig. 1 Input and output

variables of green sand AFS grain fineness no. (GFN) Permeability (P)
% of clay (% C) Green sand Green compression strength (GCS)
moulding system % of water (% W) mould system Hardness (H)
No.of strokes (NS) Bulk density (BD)

While conducting the experiments [7], sieve analysis test BD ¼ 1:02616 þ 0:01316A  0:00052B  0:06845C
was conducted to determine the grain fineness number and
size distributions of the silica sand. Moreover the strength þ 0:0083D  0:00008A2 þ 0:0009B2
of clay was obtained by performing gelling index test. þ 0:0239C 2  0:00107D2  0:00004AB
Then, the experiments were conducted with different
combinations of parameters using central composite design.  0:00018AC þ 0:00029AD  0:00302BC
Finally, the responses, such as permeability, green com-
 0:00019BD  0:00186CD: ð4Þ
pression strength, hardness and bulk density, were mea-
sured. The relationship between the responses and the The coefficient of correlation (R) is found to be equal to
process variables available in the abovementioned literature 0.9818, and the ANOVA test (refer to Table 2) was conducted
were as given below: to check the adequacy of the model is shown below.
The coefficient of correlation for permeability is found to
GCS ¼ 17:2527  1:7384A  2:7463B þ 32:3203C
be equal to 0.964. The combinations of all the linear terms,
þ 6:575D þ 0:014A2 þ 0:0945B2 that of the square terms and that of the interaction terms, have
significant contribution towards this response, as their p
 7:7857C 2  1:2079D2 þ 0:0468AB values are seen to be less than 0.05 (corresponding to 95%
 0:1215AC  0:0451AD þ 0:5516BC confidence level). It is important to mention that there is a
significant amount of lack of fit of the model after removing
þ 0:6378BD þ 2:689CD: ð1Þ the in-significant terms. Thus, the in-significant terms need
not be removed from the model. As the calculated F values
are found to be more than their respective values calculated
P ¼ 1192:51  15:98A  35:66B þ 9:51C from the table, the model is found to be statistically adequate
for making predictions. The results of the significance tests
 105:66D þ 0:07A2 þ 0:45B2  4:13C 2 for other responses also follow similar trends.
þ 4:22D2 þ 0:11AB þ 0:2AC þ 0:52AD In the present research, an attempt is being made to
optimize the process with multiple mould performance
þ 1:19BC þ 1:99BD  3:1CD: ð2Þ outputs utilizing evolutionary algorithms. A weighted
method is used for the said purpose. Since the GCS, P, H
and BD are the four different objectives, in order to
H ¼ 38:2843  0:0494A þ 2:4746B þ 7:8434C overcome the large differences in numerical values between
the objectives, the function corresponds to every mould
þ 7:774D þ 0:001A2  0:00389B2  1:6988C 2
performance output is normalized. Then weighted method
 0:6556D2  0:0015AB  0:0151AC is adapted to the normalized performance outputs to form a
single objective function. Hence, the resultant weighted
 0:0006AD  0:075BC  0:1938BD objective function to be maximized is:
þ 0:65CD: ð3Þ Maximize Z ¼ ðw1  f1 þ w2  f2 þ w3  f3 þ w4  f4 Þ ð5Þ
Subjected to constraints:
Table 1 Process parameters and their ranges 52  A  94 ð6Þ
Number Parameters Symbol Range

High Low 8  B  12 ð7Þ

1 Grain fineness number A 94 52

2 % clay content B 12 8 1:5  C  3 ð8Þ
3 % water content C 3 1.5
4 Number of strokes D 5 3
3D5 ð9Þ
12 Int J Adv Manuf Technol (2012) 58:9–17

Table 2 Results of ANOVA—

permeability Source DF Sequenced SS Adjusted SS Adjusted MS F p Ftable

Regression 14 308,592 308,592.4 22,042.3 226.4 0 1.78

Linear 4 261,699 48,629.5 12,157.4 124.9 0 2.45
Square 4 32,671 32,671.2 8,167.8 83.89 0 2.45
Interaction 6 14,222 14,222.3 2,370.4 24.35 0 2.18
Residual error 120 11,683 11,683.1 97.4
Lack of fit 10 10,815 10,815.4 1,081.5 137.1 0
Pure error 110 868 867.7 7.9
Total 134 320,275

where f1, f2, f3 and f4 are the normalized functions for 3.2 Particle swarm optimization
GCS, P, H and BD, respectively. Moreover, w1, w2, w3 and w4
are the weighted factors for the normalized GCS, P, H and PSO is a population-based stochastic optimization tech-
BD, respectively, and A, B, C and D are the process variables. nique. Due to its easy implementation and quick conver-
It is important to note that the weighted factors are selected in gence, PSO has gained much attention in solving many
such a way that their sum will be equal to one. A higher complex problems [9, 13, 14]. PSO algorithm is a model
weighing factor for an objective indicates more importance to that mimics the movement of individuals in a group. In the
that particular objective. Five different cases have been present study, MOPSO-CD [19], a variant of PSO has been
considered after choosing different values for the weights: case utilized for the selection of optimum process parameters of
1: w1 =0.25, w2 =0.25, w3 =0.25 and w4 =0.25; case 2: w1 = green sand mould system. The schematic diagram showing
0.70, w2 =0.10, w3 =0.10 and w4 =0.10; case 3: w1 =0.10, w2 = the working cycle of the PSO is shown in the Fig. 3. The
0.70, w3 =0.10 and w4 =0.10; case 4: w1 =0.10, w2 =0.10, w3 = present approach incorporates the crowding distance (that
0.70 and w4 =0.10 and case 5: w1 =0.10, w2 =0.10, w3 =0.10 is, the average distance of its two neighbouring solutions)
and w4 =0.70. These values are selected randomly in such a and mutation operators into the simple PSO algorithm. This
way that the sum of the weights will be equal to one. feature enhances the exploring capability of the algorithm
by preventing the premature convergence problem of PSO
algorithm. Instead of using evolutionary operators, such as
3 Tools and techniques used selection and crossover, each particle in the population
moves with velocity which is dynamically adjusted. The
In the present work, evolutionary algorithms, such as a new position and velocity of the particles have been
binary coded GA and PSO, have been employed to
optimize the single objective function (refer to Eq. 5) of Start
the green sand mould system. The descriptions of these
algorithms are provided in the subsequent subsections.
Randomly initialize
3.1 Genetic algorithms population of solutions
Gen = 0
Genetic algorithms are population-based search and optimiza-
tion procedures, extensively used in the search and optimiza-
tion of various problem domains [10–15]. The block diagram Is No Assign fitness to
Gen > max gen
showing the working cycle of a GA is shown in Fig. 2. The all solutions
?
selection criterion used in the present study is tournament
selection, and uniform crossover is being used as crossover Yes
Gen=gen+1 Reproduction
mechanism. Finally, bit-wise mutation is used to avoid the
local minima if any. As there are four process variables, each Stop
Crossover
variable is represented with the help of ten bits. Therefore,
40 bits are used to represent a GA string as shown below.
Mutation
1
|fflffl.{zffl
. .ffl1} 0
|fflffl.{zffl
. .ffl1} 1
|fflffl.{zffl
. .ffl0} 0|fflffl.{zffl
. .ffl0}
A B C D Fig. 2 Flow chart of the genetic algorithm
Int J Adv Manuf Technol (2012) 58:9–17 13

Start 4.1 Genetic algorithms

A parametric study (that is, by varying one parameter of

Initialize population
GA, namely probability of crossover (pc), probability of
mutation (pm), population size and maximum number of
Generate initial population generations at a time) has been conducted to determine the
combination of GA parameters that are responsible for the
Evaluate fitness optimal mould performance. It is also important to note that
the selection of the weighting factor is also important, and it
should be selected based on the requirement of the decision
Evaluate statistics
maker. In this study, five different cases (refer to Table 2)
have been considered after varying the weighing factors of
Compute crowding distance the objectives. The results of the parametric study are
shown in Fig. 4, and the procedure for conducting the
Update position & velocity
Update population systematic study is as follows.
Figure 4a shows the variation of fitness with change in the
probability of crossover (Pc) after keeping the probability of
Evaluate swarm
mutation, population size and number of generations at a
fixed level. As the problem to be solved is a maximization
Is No problem, the probability of crossover value (Pc*) which
optimal solution
obtained produced the maximum fitness has been chosen from this
study. The variation of fitness with probability of mutation
Yes (Pm) is shown in Fig. 4b. In this study, the probability of
Terminate crossover is set at Pc* and population size and number of
generations have been kept at the same level as given in
End Fig. 4a. The probability of mutation (Pm*), which is
responsible for maximum fitness is identified. Figure 4c
Fig. 3 Flowchart of the particle swarm optimization shows the variation of fitness with population size (pop). The
probability of crossover and mutation are kept at Pc* and
Pm*, respectively. Moreover, the number of generations is
calculated using the formulation given below: kept at the same level as discussed in Fig. 4a, b. In this case
also, the population size (pop*) which is responsible for
The new velocity :V ½i ¼ W  V ½i þ R1 ½Pbest ½i  P½i
ð10Þ maximum fitness has also been identified. Finally, the study
þ R2  ½AðGbest Þ  p½i has been conducted to determine the maximum number of
generations (gen*) that maximized the fitness (refer to
Fig. 4d), after fixing the other parameters, such as probability
The new position : P½ ¼ P½i þ V ½i ð11Þ of crossover, probability of mutation and population size at
Pc*, Pm* and pop*, respectively. The values of the GA
where W is the inertia weight, which is equal to 0.4, R1 and parameters obtained by this study are as given below:
R2 are the random numbers in the range of [0, −1], Pbest[i]
probability of crossover (pc*)=0.85
is the best population that the particle I reached and A(Gbest)
probability of mutation (pm*)=0.18
is the global best guide for each dominated solution. The
population size (pop*)=130
parameters, namely swarm size, number of generations,
max. number of generations (gen*)=100
inertia weight (W), social components R1 and R2 and
repository size, play an important role in the present Table 3 shows the optimum conditions of the mould
approach. parameters for multiple performance outputs with different
combinations of the weight factors. Moreover, the maxi-
mum fitness values for cases 1 to 5 are found to be equal to
4 Results and discussion 0.6654, 0.7620, 0.8522, 0.7629, and 0.6072, respectively.
Case 3 is recommended because it gives maximum green
The results of computer simulations carried out using GA compression strength, moderate permeability, maximum
and PSO are discussed below. hardness and maximum bulk density.
14 Int J Adv Manuf Technol (2012) 58:9–17

Fig. 4 GA parametric study: a

Pc vs fitness, b Pm vs fitness, c
population size vs fitness and d
maximum generations vs fitness

(a) (b)

(c) (d)
4.2 Particle swarm optimization increase computational efforts and may lead to slow
convergence. The results of this study are shown in
Here also, a systematic study has been conducted to Fig. 5. The method of conducting the systematic study is
determine the swarm size, inertia weight and maximum as follows:
number of generations. The study is conducted by varying Figure 5a shows the variation of fitness with the change
one parameter at a time. It is important to note that small in the values of inertia weight. During this process, the
swarm size may result in local convergence; large size will other two parameters, such as swarm size and generations

Table 3 Optimum mould parameters for multiple responses with different weighing factors using GA

Process parameters and responses Optimum values of mould parameters and responses

Case 1 (w1 =0.25, Case 2 (w1 =0.70, Case 3 (w1 =0.10, Case 4 (w1 =0.10, Case 5 (w1 =0.10,
w2 =0.25, w3 =0.25, w2 =0.10, w3 =0.10, w2 =0.70, w3 =0.10, w2 =0.10, w3 =0.70, w2 =0.10, w3 =0.10,
w4 =0.25) w4 =0.10) w4 =0.10) w4 =0.10) w4 =0.70)

A: GFN 52.0007 52.0012 93.9998 93.9982 93.9999

B: %C 11.9997 8.0001 11.9999 11.9997 11.9998
C: %W 2.9886 2.4883 2.6546 2.9999 2.4601
D: NS 4.9999 3.0000 4.9998 4.9994 4.9999
GCS 55.5022 21.2056 54.9377 53.8732 54.7159
P 107.1197 214.501 53.6790 54.9672 52.5185
H 84.9847 77.3861 89.4473 86.9985 86.1361
BD 1.5044 1.4726 1.5888 1.5799 1.5938
Int J Adv Manuf Technol (2012) 58:9–17 15

Fig. 5 PSO parametric study: a

inertia weight vs fitness, b
swarm size vs fitness and c
maximum generations vs fitness

(a) (b)

(c)

are kept at the fixed level. In this case, the inertia weight In this study, the swarm size (SS*) that is responsible for
value (W*), corresponding to the maximum fitness is maximum fitness has been found. Figure 5c shows the
identified. Figure 5b shows the study related to the swarm convergence of the solution over number of generations.
size. In this case, inertia weight is set at W* and maximum The number of generations (G*) that are responsible for
generations are set at the same level as discussed in Fig. 5a. maximum fitness has been obtained in this study. Thus, the

Table 4 Optimum mould parameters for multiple responses with different weighing factors using PSO

Process parameters and responses Optimum values of mould parameters and responses

Case 1 (w1 =0.25, Case 2 (w1 =0.70, Case 3 (w1 =0.10, Case 4 (w1 =0.10, Case 5 (w1 =0.10,
w2 =0.25, w3 =0.25, w2 =0.10, w3 =0.10, w2 =0.70, w3 =0.10, w2 =0.10, w3 =0.70, w2 =0.10, w3 =0.10,
w4 =0.25) w4 =0.10) w4 =0.10) w4 =0.10) w4 =0.70)

A: GFN 52.0024 52.0158 52.0001 93.9935 93.9994

B: %C 11.9999 8.0002 11.9998 11.9987 11.9977
C: %W 2.9042 2.6698 2.8452 2.9999 2.3084
D: NS 4.9937 3.0000 4.9999 5.0000 4.9999
GCS 55.4251 20.8928 55.4112 53.8727 54.1248
P 107.7204 214.2018 107.8949 54.9601 51.4005
H 84.8603 77.6046 84.7936 86.9991 85.8925
BD 1.5065 1.4701 1.5079 1.5799 1.5976
16 Int J Adv Manuf Technol (2012) 58:9–17

parameters of PSO that are responsible for the better that PSO has performed better than GA in terms of
performance are as follows: computational efficiency. Moreover, the percent deviation
with the experimental results for all the responses for PSO
inertia weight (W*)=0.2
is less than that of the GA. The simple structure associated
swarm size (SS*)=50
with minimal parameter tuning helps the PSO in out-
number of generations (G*)=30
performing the GA.
In this case also, five different cases as mentioned in the
above approach have been considered after varying the
weight factors of the objectives. Table 4 shows the
References
optimum conditions of the mould parameters for multiple
performance outputs with different combinations of the
weight factors. It is interesting to note that the maximum 1. Marek CT (1966) Green sand permeability—its significance and
control. AFS Trans 74:70–81
fitness values for cases 1 to 5 are found to be equal to 2. Frost J, Hiller JM (1966) The mechanics of green sand moulding.
0.6647, 0.7609, 0.8719, 0.7630 and 0.6066, respectively. In AFS Trans 74:177–186
this method also, case 3 is recommended, as it has produced 3. Wenninger EC (1968) Green sand processing: an introduction to
maximum fitness. current fundamental thinking. AFS Trans 76:101–111
4. Dhindaw KB, Chakraborty M (1974) Study and control of
properties and behavior of different sand systems by application
4.3 Comparison between GA and PSO of statistical design of experiments. In: Proceedings of the 41st
international foundry congress, pp 9–14
While applying the evolutionary algorithms, such as GA 5. Casalino G, De Filippis LAC, Ludovico A (2005) A technical
note on the mathematical and physical characterization of
and PSO, to optimize a particular system, a number of selective laser sintered sand for rapid casting. J Mater Process
parameters are required to be specified. The speed of Technol 166:1–8
convergence of the algorithm depends on an appropriate 6. Parappagoudar MB, Pratihar DK, Datta GL (2007) Linear and
choice of the parameters. The optimal parameters of GA are non-linear statistical modeling of green sand mould system. Int J
Cast Met Res 20(1):1–13
found to be equal to probability of crossover 0.85,
7. Parappagoudar MB, Pratihar DK, Datta GL (2007) Non-linear
probability of mutation 0.18, population size 130 and modeling using central composite design to predict green sand
maximum number of generations 100. In the case of PSO, mould properties. Proc IMechE B J Eng Manufacture 221:881–
the optimal parameters of inertia weight, swarm size and 894
8. Sushil kumar, Satsangi PS, Prajapati DR (2010) Optimization of
number of generations are found to be equal to 0.2, 30 and
green sand casting process parameters of a foundry by using Taguchi
50, respectively. It is clear that the convergence rate of PSO method. Int J Adv Manuf Tech. doi: 10.1007/s00170-010-3029-0
is faster than that of the GA. Moreover, confirmation 9. Rosenberg RS (1967) Simulation of genetic populations with
experiments were conducted for the case 3 of both the GA biochemical properties. Ph.D. thesis, University of Michigan
10. Lee TS, Ting TO, Lin YJ (2007) An investigation of grinding
and PSO. The percentage errors associated with GCS, P, H
process optimization via evolutionary algorithms. In: Proceedings
and BD are found to be equal to 3.25%, 8.46%, 5.35% and of the 2007 IEEE swarm intelligence symposium
2.45%, respectively, for GA, and 3.22%, 5.78%, 4.95% and 11. Sardinas RQ, Santana MR, Brindis EA (2006) Genetic algorithm-
2.25%, respectively, for PSO. Moreover, the CPU times based multi-objective optimization of cutting parameters in
required by GA and PSO, for a population size of 50 and turning process. Eng Appl Artif Intell 19:127–133
12. Datta R, Majumder A (2010) Optimization of turning process
number of generations equal to 30, are found to be equal to parameters using multi-objective evolutionary algorithms. In:
0.021 and 0.013 s, respectively, on a P-IV machine. It is Proceedings of the IEEE congress on evolutionary computation,
interesting to note that PSO has performed better than GA Barcelona, pp 1–6
for all the responses. This may be due to the simple 13. Ali-Tavoli M, Nariman-Zadeh N, Khakhali A, Mehran M (2006)
Multi-objective optimization of abrasive flow machining process
structure and minimal parameter tuning of PSO compared using polynomial neural networks and genetic algorithms.
to GA. Machining Sci Technol 10(4):491–510
14. Kuriakose S, Shanmugam MS (2005) Multi-objective optimiza-
tion of wire-electro discharge machining process by non-
dominated sorting genetic algorithm. J Mater Process Technol
5 Conclusions
170:133–141
15. Hong X, Yuan Li, Kaifu Z, Jianfeng Y, Zhenxing L, Jianbin S
In the present work, an attempt has been made to search for (2010) Multi-objective optimization method for automatic
the optimum process parameter values for the multiple drilling and riveting sequence planning. Chin J Aeronaut
23:734–742
objectives, namely green compression strength, permeabil-
16. Rao RV, Pawar PJ, Shankar R (2008) Multi-objective optimization of
ity, hardness and bulk density, utilizing evolutionary electrochemical machining process parameters using a particle swarm
algorithms, such as GA and PSO. It is interesting to note optimization. Proc IMechE B J Eng Manufacture 222:949–958
Int J Adv Manuf Technol (2012) 58:9–17 17

17. Kumar CA, Nair NK (2010) “Multi-objective PID controller symposium 2003 (SIS 2003), Indianapolis, Indiana, USA, pp.
based on adaptive weighted PSO with application to steam 243–246
temperature control in boilers”. Int J Eng Sci Technol 2 19. Raquel C, Naval P (2005) An effective use of crowding distance
(7):3179–3184 in multiobjective particle swarm optimization. In: Proceedings of
18. Hu X, Eberhart RC, Shi Y (2003) Engineering optimization with the 2005 conference on genetic and evolutionary computation,
particle swarm. In: Proceedings of the IEEE swarm intelligence Washington DC, USA, pp 257–264